getting black outputs
nothing i do is fixing this, running the workflow provided entirely.
If you have Sage-attention installed try using the "patch sage attention" node from Kijai immediately after the checkpoint or diffusion model loader.
If you have Sage-attention installed try using the "patch sage attention" node from Kijai immediately after the checkpoint or diffusion model loader.
I run Comfy with --use-sage-attention and no black outputs at all.
I solved the same issue by replace weigth_dtype from default to fp8_e5m2 in the Load diffusion Model node
nvm im an idiot, i fixed it
nvm im an idiot, i fixed it
how ? I have the same issue
my randomly generates black images sometimes, i want to know how to fix it?
try switching to a fp8 version of the model (or gguf quant). bf16 can be unstable, causing black outputs.
i fixed by changing some code i node.py. im running on a rtx 5080 so running fp8 or gguf sound like downgrade from full model.
https://www.reddit.com/r/comfyui/comments/1p9i300/getting_only_black_images_when_running_zimage/
some in that post found a solution and it worked for me, by thanks
nvm im an idiot, i fixed it
how ? I have the same issue
i had to run the workflow in fp8 dtype with the bf16 model loaded in the node. but then it randomly fixed itself when i restarted my pc back up the next day. i simply loaded the bf16 model into the node and set it back to default and it worked fine. im not sure what the issue really was in the first place. im guessing with the fresh update, the new models, etc probably was causing bugs. the application isnt perfect either, remember!